You are on page 1of 18

Computers and Geotechnics 155 (2023) 105222

Contents lists available at ScienceDirect

Computers and Geotechnics


journal homepage: www.elsevier.com/locate/compgeo

A multiaxial inertial macroelement for deep foundations


Davide Noè Gorini *, Luigi Callisto
Sapienza University of Rome, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: A hyperplastic macroelement is proposed as an efficient means for simulating nonlinear and multi-directional
Soil-piles interaction soil-piles interaction in the nonlinear analysis of structures. The macroelement relates the generalised forces
Macroelement exchanged between the superstructure and the foundation to the corresponding displacements and rotations of
Hyperplasticity
the foundation raft. The inertial effects developing under dynamic loading can be reproduced by coupling the
Nonlinear behaviour
Frequency dependency
macroelement with the participating masses of the soil-foundation system, obtainable with a modal identification
OpenSees of the pile group. Failure conditions of the foundation are described by a hyper-ovoidal ultimate limit state
surface in the force space, which is calibrated through standardised procedures. The plastic behaviour, controlled
by a series of yield surfaces with kinematic hardening, produces the desired directional coupling and allows to
model the cyclic response and the irreversible deformation of the foundation. The macroelement is implemented
in the analysis framework OpenSees for a prompt use in earthquake engineering applications. Its predictions
under monotonic loads are validated against experimental data and advanced, fully coupled numerical model­
ling, while its dynamic response is tested on a reference case study.

1. Introduction degree of freedom of the foundation (Dobry et al., 1982; Kaynia and
Kausel, 1982; Gazetas and Dobry, 1984a, 1984b; Gazetas, 1991; Mylo­
Foundations resting on pile groups are used not only to cope with nakis and Gazetas, 1999; Mylonakis and Roumbas, 2001; Karatzia and
insufficient bearing capacity and stiffness of a parent shallow founda­ Mylonakis, 2012). This technique does not allow for the evaluation of
tion, but also to absorb eccentric and inclined loads. Therefore, these the seismic performance of the foundation, as it neglects the directional
foundations need to be designed to support several load combinations, coupling and the nonlinear features of the response, as well as the ul­
including vertical and horizontal forces and moments, that is typically timate capacity of the soil-foundation system. On the other hand, the
the case of the foundations of bridge piers. For each loading combina­ development of a full soil-structure numerical model in which the soil is
tion, the design should assess the bearing capacity (ultimate limit states) modelled as a continuum is in most cases unfeasible, especially for time-
and predict the deformation (serviceability limit states). Under seismic domain dynamic analyses.
loading, the seismic performance of the foundation must be assessed, In a macroelement approach (Nova and Montrasio, 1991; Roscoe and
that involves the time-dependent multiaxial response of a pile group. Schofield, 1956; Cremer et al., 2002; Gorini and Callisto, 2020, 2022b),
In a traditional design approach, each of these design verifications is the response of the soil-foundation system is described through a multi-
carried out separately for the vertical and the horizontal loads. By directional constitutive relationship between the generalised forces
contrast, recent studies demonstrated that the ultimate capacity of a exchanged at the structure-foundation contact and the corresponding
piled foundation is profoundly affected by the simultaneous presence of displacements and rotations. This is a very efficient technique, because it
these loads and moments (Di Laora et al., 2019; Gorini and Callisto, implies the addition of only six degrees of freedom (three displacements
2022b; Correia, 2011; Gerolymos et al., 2015). Notwithstanding, the and three rotations) to the numerical model of the superstructure. If the
seismic performance is commonly assessed through conventional ap­ model is developed in the context of hardening plasticity, the directional
proaches that describe the effect of the seismic actions in terms of coupling is a direct result of the plastic flow rule, and the seismic per­
equivalent static forces transmitted to the foundation. In a time-domain formance of the foundation is provided by the maximum and permanent
analysis, elastic springs and dashpots can be introduced at the base of a deformations due to the seismic shaking. Over the years, a great effort
structural numerical model to simulate the soil compliance for each has been dedicated to the description of the multiaxial response of

* Corresponding author at: Dept. of Structural and Geotechnical Engineering, via Eudossiana 18, 00184, Roma, Italy.
E-mail address: davideno.gorini@uniroma1.it (D.N. Gorini).

https://doi.org/10.1016/j.compgeo.2022.105222
Received 27 September 2022; Received in revised form 5 December 2022; Accepted 24 December 2022
Available online 11 January 2023
0266-352X/© 2022 Elsevier Ltd. All rights reserved.
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

shallow foundations (Chatzigogos et al., 2011; Cremer et al., 2001; Di common features of the cyclic behaviour of geotechnical systems, such
Prisco et al., 2003; Le Pape and Sieffert, 2001; Paolucci, 1997; Rha and as a linear elastic response at small-strain levels, a plasticity-based
Taciroglu, 2007; Salciarini and Tamagnini, 2009; Venanzi et al., 2014) description of irreversible effects and a kinematic hardening of the
and a multiaxial relationship was recently developed by Gorini et al. yield surfaces. In contrast, the dissipation function depends on the
(2023) within a rigorous thermodynamic framework to simulate the specific plastic domain adopted for the foundation type under
combined nonlinear and frequency-dependent response exhibited by consideration.
bridge abutments under dynamic loading. Macroelement representa­ The TIMs are aimed at simulating salient features of the dynamic
tions were also proposed to simulate the nonlinear, horizontal-vertical response of geotechnical systems, that are i) a combined nonlinear and
response of single piles (Boulanger et al., 1990; El Naggar and Bent­ frequency-dependent multiaxial response and ii) the strong dependence
ley, 2000; Houlsby et al., 2017). of the response on the load direction. Primary assumptions of the present
The present paper proposes a macroelement, formulated in the formulation are the validity of the orthogonality principle (Ziegler,
context of hyper-plasticity, that can be used to model the mechanical 1977), the additive decomposition of elastic and plastic components of
behaviour of piled foundations subjected to general loading conditions. deformations, and the associativity of the plastic flow.
The macroelement is expressly conceived for foundations with a rela­ The use of the TIMs in the structural analysis follows the procedure
tively small number of piles connected by a rigid cap, as those usually described in Gorini (2019) and Gorini et al. (2022a), illustrated in Fig. 1
employed for bridges. The model is based on the failure surface for pile for the case of a bridge: the seismic motion is propagated through the
groups proposed by Gorini and Callisto, 2022a) and the methodology foundation soils by means of a free-field site response analysis; the
extends the thermodynamic-based formulation developed by Gorini motion obtained from this analysis at an effective depth from the
et al. (2022a, 2023) for the case of bridge abutments. foundation represents the seismic input for the TIM, which is included in
In the following sections, the analytical formulation of the proposed the global structural model to carry out time-domain nonlinear dynamic
macroelement is described in detail, its monotonic response is validated analyses of the soil-structure system.
against recently published experimental data and advanced numerical The use of TIMs is quite immediate for bridge foundations and
results, and its dynamic response is critically discussed, also with abutments, but the same analysis procedure can also be extended to the
reference to the case study of a specific bridge. case of buildings with isolated foundations. The remarkable computa­
tional efficiency of this class of models, comparable with the one asso­
2. The Tim approach ciated with classical substructure linear approaches, was exhaustively
discussed in Gorini et al. (2022a, 2023) and Gorini and Callisto (2022b).
The macroelement for piled foundations is part of a novel analysis
method for studying dynamic soil-structure interaction, called TIM 3. Formulation
approach (Gorini et al., 2022a, 2023; Gorini and Callisto, 2022b). In this
method, Thermodynamic Inertial Macroelements (TIMs) are used to The macroelement for the soil-piles system, called SPME, is illus­
simulate the response of geotechnical systems in the assessment of trated with reference to the schematic foundation layout depicted in
structures, accounting for nonlinear and frequency-dependent features Fig. 2a. It is conceived as a multiaxial, nonlinear relationship between
under multiaxial loading. The first reference for this class of macroele­ the generalised forces, Qi, exchanged between the superstructure and
ments is constituted by the model proposed by Gorini et al. (2023) for the foundation slab, and the corresponding displacements and rotations,
bridge abutments. In general, a TIM consists of a relationship between qi, considering the directions i as in Fig. 2a (i = 1, 2 refers to horizontal
the generalized forces exchanged at the soil-structure contact and the translations, i = 3 is the vertical translation, i = R1, R2 are the rotations
corresponding displacements and rotations. It is formulated as a multi- around axis 1 and 2, respectively). The SPME encloses the mechanical
surface constitutive law with kinematic hardening derived within a response of the piles and of the soil interacting with it. The force­
rigorous thermodynamic framework, using hyper-plasticity (Collins and –displacement relationship reads:
Houlsby, 1997). The macroelement response is completely defined by
Q̇i = Hij • q̇j ; i = 1, 2, 3, R1, R2; j = 1, 2, 3, R1, R2 (1)
two potentials, namely the energy and dissipation functions. Different
TIMs have the same energy function since this function incorporates

Fig. 1. Application of the macroelements of the TIM approach in the structural analysis of a bridge.

2
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

Fig. 2. A) Schematic layout of the generalised forces exchanged between the superstructure and the piled foundation in a central vertical section; b-d) fitting of the
numerical results (filled circles) with the proposed analytical model of ultimate surface (reproduced by Gorini and Callisto, 2022a).

in which Hij is the second-order tangent stiffness matrix. The contribu­ are bounded by the ultimate limit state surface (n = N), which is the locus
tion of the external moment QR3 around the vertical axis is neglected. in the force space corresponding to the attainment of the ultimate capacity
of the foundation. For the present development, the ultimate surface is
described by the relationship proposed by Gorini and Callisto, 2022a). This
3.1. Plastic domain of the macroelement is depicted in Fig. 2b,c,d for the example piled foundation considered in
that study, with reference to loads in the 1–3 space only. Note that the
The dissipative response of the SPME is controlled by N yield surfaces, external moment QR2 is divided by the width of the raft in direction 1, B1.
which compose the so-called plastic domain. Within the surface of first The analytical expression of the ultimate locus is:
yield (n = 1) the response is linearly elastic, while the interaction forces

( ) ( ) ]− 1
( ) [ ( ) ̂3
A ̂3
A ( )
̂ 3,1 2 • A
y(N) = 2 • Q1 • Q3 + Q1 • Q ̂3 • ̂ ̂ 2 • ̂c 3 + Q
S F,1 • Γ ̂ 3,1 2 • − Q3 + +̂ c 3 • Q3 + c3
− ̂ ̂ 3,2 2
+ 2 • Q2 • Q3 + Q2 • Q
1
2 2
[ ( ̂3 ) ( ̂3 ) ]− 1
( )2 A A
• Â3 • ̂ ̂2 • ̂
S F,2 • Γ c 3 + ̂
Q 3,2 • − Q 3 + + c
̂ 3 • Q3 + − c
̂ 3 − 1= 0 (2)
2
2 2

3
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

in which the moment components QR1 and QR2 are included in the set of (l) (l)
{ } g = Qi • qi − f
evolution functions ̂c 3 , Q
̂ 3,1− 2 , A S F,1 − 2, Γ
̂ 3, ̂ ̂ 1− 2 , defined as:
1 ∑N
1 ∑N
[ ( )nc,3 ]n 1 = − • Cij(0) • Q(0) (0)
j • Qi − Q(n) (n)
i • qi + • H (n) • q(n) (n)
j • qi (9)
|QR2 | c,3
2 2 n=1 ij
c 3 = c(0)
̂ 3 + ac,3 • 1 − (3) n=1
Q(max) • bc,3
where H(0) (0)
R2
ij and Cij are the elastic stiffness and compliance matrices and
[ ( )nQ3,1− 2 ]n qi represents, when k = 0, the elastic displacement in the i-direction
1 (k)
Q3,1− 2
|QR2− R1 |
̂ 3,1−
Q 2 = aQ3,1− 2 • 1 − (4) and, and for k = 1,2,…,N, the kth plastic displacement. The part of g
Q(max)
R2− R1 • bQ3,1− 2
( )
depending only on the internal variables, g2 qi
(n)
=
[ ( )nA3 ]n 1 ∑N
1/2 • n=1 Hij • qj • qi , encloses the contribution of the kinematic
(n) (n) (n)
A3
|QR2− R1 |
̂
A 3 = aA3 • 1 − (max)
(5) hardening associated with the yield surfaces.
QR2− R1 • bA3
The dissipation function d is instead specific of the SPME formulation
[ ( )nSF,1− 2 ]n 1 since it is a function of the shape of the yield surfaces under the
assumption of associated plastic flows. It is given by the following
SF,1− 2
|QR2− R1 |
̂
S F,1− 2
(0)
= SF,1− 2 + aSF,1− 2 • 1 − (6)
expression:
(max)
QR2− R1 • bSF,1− 2
( )
[ ( )nΓ,1− 2 ]n 1 ∂y(n)
(10)
(n)
|QR2− R1 | Γ,1− 2 d = y(n) • χ i • (n) − y(n)
̂
Γ 1− 2 = aΓ,1− 2 • 1 − (max)
(7) ∂χ i
QR2− R1 • bΓ,1− 2
in which χ i is the dissipative force vector of the nth yield
(n) (n)
Equations (2) to (7) describe a hyper-egg with super-elliptical = ∂d/∂q̇i
generatrices in the five-dimensional force space. surface, that coincides with the nth generalised force vector ¯χ i
(n)
= ∂g/
The ovoidal shape of the failure locus in Q1-Q3 space is produced by ∂q(n) for the validity of the orthogonality principle (Ziegler, 1977), and
the variability of the yield moment of the pile cross-section with the
i
λ(n) ≥ 0 is the nth plastic multiplier derived in the following. It can be
relative axial force. The presence of the external moments acting on the
demonstrated that the true forces, Q(n) i , are related to the dissipative
foundation always reduces the combined horizontal-vertical limit load
forces as Qi = χi + ci , where c(n) th
i (centre of the n yield surface)
(n) (n) (n)
according to super-elliptical relationships (Eqs. (3)-(7)). Each super-
represents the so-called back force due to kinematic hardening (see, for
ellipse is a function of the mobilised moment |QR2− acting
(max)
R1 |/QR2− R1
instance, Gorini, 2019, or Gorini et al., 2023).
in the respective plane (QR2− R1
is the maximum moment that can be
(max)

applied to the foundation, see Gorini and Callisto, 2022a) and is 3.2.2. Incremental response
completely defined by four parameters: G(0) 1-2, aG,1-2, bG,1-2 and nG,1-2, The relationship between the rates of the external forces and of the
where G indicates the generic parameter of the evolution function, deformations can be obtained by differentiating the Helmholtz free en­
whose calibration is recalled in Section 5. ergy as follows:
As discussed by Gorini et al. (2023) for the case of bridge abutments, ( ) ( )
the internal yield surfaces are assumed to be homothetic to y(N), ∑N
∂ ∂f
(11)
(0) (n)
Q̇i = = H ij • q̇j − q̇j
assuming a linear decrease of their size down to the surface of first yield, ∂t ∂qi n=1
the latter having typically a size of 10 % of the ultimate locus.
whose solution requires the definition of the flow rule for each yield
3.2. Thermodynamic framework function, associated by hypothesis:

The thermodynamic-based incremental response of the SPME is


∂y(n)
q̇(n)
i = y(n) • (12)
completely derived by the energy and dissipation potential functions, ∂χ (n)
i

that are consistent with the First and Second Laws of Thermodynamics
(Collins and Houlsby, 1997; Houlsby and Puzrin, 2006). The plastic in which, after some manipulation (full derivation in Appendix), the
displacements, q(n) gradient of the nth yield surface assumes the following form:
i (n = 1,…, N), play the role of internal variables and

the total displacement in the i-direction is equal to qi = qi + N n=1 qi ,
( ) ( )
(0) (n)
∂y(n)
(13)
(n) (n) ̂ 3,1 • χ(n) (n) (n) ̂
= N1 = 4 • χ3 + Q 1 • χ3 + χ1 • Q 3,1
where qi is the elastic displacement.
(0) (n)
∂χ1

( ) ( )
3.2.1. Thermodynamic-based potentials ∂y(n)
(14)
(n) (n) ̂ 3,2 • χ(n) (n) (n) ̂
= N2 = 4 • χ3 + Q 2 • χ3 + χ2 • Q 3,2
The energy function represents the mechanical work done by the (n)
∂χ2
forces acting in the system and is conveniently expressed by the Helm­
holtz free energy, f, or equivalently by the Gibbs free energy, g. The TIMs ∂y(n)
= N3(n)
have the same energy function (Gorini and Callisto, 2022b), that is: ∂χ(n)
3
( ) ( ) ( ) (
1 ∑N ∑N
1 ∑ N ∂y(n) ∂y(n) ∂y(n) ∂y(n) ∂y(n)
f = • Hij(0) • qj − q(n) • qi − q (n)
+ • Hij(n) • q(n) (n)
j • qi
= N1
(n)
− F1 • D1
(n)
• y(n) (n)
D2 + yD1 •
D2
+ N2 − F2 • D3

2 n=1
j
n=1
i
2 n=1
∂χ3 ∂χ3 ∂χ(n)
3
(n)
∂χ3 ∂χ(n)
3
)
(8) ∂y(n)
(15)
(n) (n)
• yD4 + yD3 • D4
∂χ(n)
from which one can obtain the Gibbs free energy by using the Legendre 3

transform, such that:


∂y(n) ∂y(n) ∂y(n)
(n)
(n)
= NR1 = N2 − D5
(16)
∂χR1 ∂χ(n)
R1 ∂χ(n)
R1

4
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

∂y(n) ∂y(n) ∂y(n) 5. Calibration Procedure


(n)
(n)
= NR2 = N1 − D6
(17)
∂χR2 ∂χ(n) ∂χ(n)
The calibration of the proposed macroelement for piled foundations
R2 R2

When the current equilibrium state Qi is within the nth yield surface, requires the definition of 1) the ultimate limit state surface, 2) the initial
this surface does not contribute to the hardening response; by contrast, stiffness matrix and 3) the participating masses of the soil-foundation
when Qi lies on the nth yield surface, plastic deformation can occur as a system.
function of the respective plastic multiplier, λ(n) > 0 (plastic loading).
The latter is determined by invoking the consistency condition (ẏ(n) = 0) 5.1. Ultimate limit state surface
and, in the thermodynamic framework, Houlsby and Puzrin (2006)
demonstrated that it assumes the following form: By virtue of the homotheticity of the yield surfaces, the plastic
(n) (n)
domain is defined by the sole identification of the ultimate yield surface,
∂y(n) ∂y(n)
(n)
∂χ i
• Q̇i (n)
∂χ i
• Q̇i that is in turn based on the evaluation of the vertical and horizontal limit
loads of the pile group, the maximum external moment (Di Laora et al.,
(n)
y = ∂y(n) ∂2 g2 (n) ∂y(n) (n)
= ∂y(n) ∂2 g2 (n)
, n = 1, 2, ⋯, N
• • ∂y(n) − • ∂y(n) • • ∂y(n)
∂χ (n)
i
∂q(n)
i
∂q(n)
j
∂χ j ∂q(n)
i
∂χ i ∂χ (n)
i
∂q(n)
i
∂q(n)
j
∂χ j 2019; Gorini and Callisto, 2022a) and of the parameters Sf0, aG and aQ3
(18) controlling the shape of the hyper-egg (see Section 3.1). The latter ones
can be estimated expeditiously through the non-dimensional relation­
in which the term ∂y(n) /∂qi = 0 because the yield functions depend only
(n)
ships proposed by Gorini and Callisto, 2022a) or, for a more specific
on the dissipative forces, χ(n)
i , and not on the plastic displacements. The assessment, by using numerical solutions on the failure loads of a pile
second derivative of g2 is exactly the nth stiffness matrix Hij , repre­
(n) group under general loading. In this regard, one can for instance refer to
the code DeepFUL (Gorini and Callisto, 2022a) providing i) a real-time,
senting the contribution of the kinematic hardening to the dissipative
conservative evaluation of the generalised failure loads for a pile group
response of the SPME. The denominator of λ (n) represents the nth plastic
and ii) the calibration of the five-dimensional, ovoidal model of ultimate
modulus which is therefore equal to ∂y(n) /∂χi • Hij • ∂y(n) /∂χj .
(n) (n) (n)
surface.
During plastic loading, the yield surfaces instantly reached by the The choice of the number of yield surfaces depends on the desired
force vector translate so that ẏ(n) = 0, keeping their shape and size un­ smoothness for the piecewise linear force–displacement law of the
changed. Within the hyperplastic framework, the relative translation SPME. As it will be shown later, a limited number of surfaces, generally
rule derives from the energy function, guaranteeing the thermodynamic not greater than 10, leads to a sufficient accuracy of the response. It is
consistency of the deformation process. Since the TIMs have the same worth noticing, however, that this choice has a minor influence on the
energy function, the hardening law is exactly the one derived for the computational effort, as it increases only the iterations in the material
case of bridge abutments (Gorini et al., 2022a, 2023), based on which sub-routine but does not alter the number of degrees of freedom in the
the evolution law for the centre of the nth yield surface reads ċi
(n)
= numerical model.
Hij q̇i .
(n) (n)

5.2. Initial stiffness and participating masses

4. Implementation and Use


The participating masses of the soil-foundation system along the five
degrees of freedom of the SPME are calibrated to reproduce the dynamic
The SPME was implemented in the finite element analysis framework
amplification of the response at the small-strain resonance periods of the
OpenSees (McKenna, 1997; McKenna et al., 2000) as the extension of the
system. Each mass component mi is calculated as a function of the small-
NDMaterial subclass, named NDTIMaterials (Gorini et al., 2023). The
strain stiffness H(0) (0)
ij and of the fundamental vibration period Ti at small
new C++ source code includes the multiaxial constitutive relationship
strains. Preliminary investigations on the directional coupling of the
described above, that can be assigned to a novel zero-length finite
SPME response, here omitted for the sake of brevity, showed that it is
element, ZeroLength6D, developed in the present study, establishing an
dominated by the development of plastic displacements since the SPME
explicit nonlinear relationship between two overlapped nodes. This
is a multi-yield model with a narrow elastic core. Accordingly, for
element extends the one available in the OpenSees library, Zer­
simplicity the stiffness matrices H(l) ij (l = 0,1,…,N) are assumed to be
oLengthND, working on the translational degrees of freedom of the nodes
diagonal, hence the directional coupling of the macroelement response
only, to a fully coupled translational-rotational response.
is provided by the plastic flow rule only. The stiffness matrix H(1) ij ,
The initial state of the macroelement represents the end of the con­
controlling the kinematic hardening of the first yield surface, is assumed
struction stage for the foundation, corresponding to Qi = 0 (superstruc­
equal to H(0)
ij to reproduce the period lengthening due to nonlinear soil
ture not built yet). The effects of different installation technologies of the
behaviour starting from a mobilised strength Qi / Q(max) i > 30 % (Gorini
piles are incorporated in the calibration of the vertical and horizontal
et al., 2022a). The values of H(0) ij and Ti
(0)
can be identified through
ultimate capacity of the pile group (input parameters for the plastic
conventional solutions for the impedance functions of pile groups or
domain) and of the initial stiffness matrix through standard methodol­
carrying out a linear analysis of the foundation, in which the eventual
ogies (Viggiani et al., 2012). During the static stages that represent the
mutual interaction between the piles can be accounted for using
construction sequence of the bridge, the end node of the macroelement
standardised methodologies. For instance, the dynamic identification of
is fully restrained. In the subsequent seismic stage, the SPME is com­
a pile group can be straightforwardly accomplished by using the
bined with the participating masses of the soil-foundation system, mi.
software DYNA6 (https://www.eng.uwo.ca/grc/dyna6/index.html),
These masses are assigned to the generic base nodes of the global
providing the evolution of the dynamic stiffness of a pile group with the
structural model to which the SPME is connected. The restraint at the
period of the loading function. The initial stiffness, H(0) (1)
ij = Hij , is
end node of the SPME is removed and the seismic motion is applied to
computed as the dynamic stiffness at large periods, while the funda­
this node. As a result of a sensitivity analysis, omitted for brevity, it was
mental vibration period T(0) i corresponds to the minimum of the
seen that the seismic motion can be obtained as the one computed by a
impedance function. Once H(0) ij and Ti
(0)
are known, the participating
free-field site response analysis at a depth equal to 10D, where D is the
masses of the soil-piles system are computed as m(SP) i = 8π2 H(0) (0)2
ij × Tj .
pile diameter.
The participating mass of the entire foundation system is then given by
mi = m(SP)
i + m(raft)
i , where m(raft)
i is the mass, for the translational degrees
of freedom, or rotational inertia, for the rotations, of the raft assumed
entirely participating to the dynamic response.

5
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

6. Validation under monotonic loading conditions MN. The vertical capacity in tension was derived from the experimental
measurement of the tension limit load for a single pile, using the same
In this section, the multiaxial response of the SPME is validated pile group efficiency estimated from the experiments for the compres­
against experimental and numerical results available in the literature for sive capacity. This allowed to estimate a bearing capacity of the group in
some well-documented case studies. tension as Q(-)3 = − 1.5 MN. As per the maximum moment capacity,
Q(max)
R , Gorini and Callisto, 2022a) found that this may be taken to occur
6.1. Static vertical loads (case study No. 1) for a critical value of the vertical load Q(cr) (-) (+)
3 = (Q3 + Q3 ) / 2, for which
an increase of the applied moment produces a plastic mechanism in
The first case study includes a series of centrifuge experiments (de which half of the piles reach the vertical capacity in compression and
Sanctis et al., 2021) in which the behaviour of pile groups under centred half the capacity in tension. In these conditions, the maximum external
and eccentric vertical loads was investigated. Two sets of tests were moment around the generic horizontal axis i can be computed through
performed at 50g on annular-shaped pile groups consisting of 8 the following analytical solution (Gorini and Callisto, 2022a):
aluminium piles, and on isolated single piles embedded in kaolin clay. ∑
N
Different eccentricities of the vertical force with respect to the centre of Q(max)
R = Q(cr)
3 • ih • 〈nh − (2j − 1)〉 • nk (19)
the raft were considered. Fig. 3a,b illustrate the arrangement of the j=1

foundations in the cylindrical container of the centrifuge related to the


where the subscript h denotes the horizontal direction orthogonal to
so-called set A of the experimental program (dimensions refer to the
direction k, ih is the pile spacing in the h-direction and the symbols 〈.〉 are
model scale whereas the values in parentheses to the prototype di­
the Macaulay brackets.
mensions). The simulation of this experiment is carried out at the pro­
The ultimate surface obtained for this experiment is depicted in
totype scale: the piles have an outer diameter of 0.5 m and a length of 14
Fig. 3c in the homogenised force space Q3 - QR/R, with R = 3 m being the
m, and are connected with spherical hinges to an aluminium circular raft
radius of the annular pile group. It can be seen that the analytical model
not in contact with the underlying soil, having an outer diameter of 6.9
of ultimate surface passes through the three input combined loads of
m. A vertical displacement is enforced at the tip of a cantilever beam
coordinates (0, Q(-) (+)
3 ), (0, Q3 ) and (QR
(max)
/ R, Q(cr)
3 ).
attached to the cap, producing a resulting vertical load with a constant
The initial vertical stiffness of the SPME, H(0)33 , was determined from
eccentricity.
the experimental force–displacement relationship, discussed in the
Macroelement predictions were obtained for all the loading paths
following section, while the rotational initial stiffness H(0) RR was derived
investigated in the experimental program. The macroelement was cali­
by assuming an identical scale factor of the initial stiffness and of the
brated along the lines illustrated in the previous section, as follows. The
ultimate capacity in the force space Q3 - QR/R, such that:
ultimate locus in the Q3 - QR force space (vertical force-moment) re­
quires the determination of the bearing capacity of the pile group under (1D)
R • QRR
compression, Q(+) (-)
3 , and tension, Q3 , and the maximum moment that can
(0)
HRR (0)
= H33 • (20)
(max) Q(+)
be applied to the foundation, QR . The vertical capacity in compres­ 3

sion was taken directly from the experimental data, equal to Q(+) 3 = 2.5

Fig. 3. A) Vertical section and (b) plan view of the piled foundations of set A (dimensions refer to the prototype scale) analysed in the geotechnical centrifuge by de
Sanctis et al. (2021); c) failure locus of the macroelement in the Q3-QRi/R space for case A2.

6
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

Fig. 4. Validation of the macroelement under monotonic vertical loads: a) q3-Q3 response under a centred load, b) qR1R-Q3/R responses under vertical loads with
high and small eccentricity (cases A2 and B1, respectively).

where Q(1D)
RR is the rotational capacity of the pile group when Q3 = 0. The final part of the B1 experimental curve is quite irregular.
stiffness associated with combined load paths is then a result of the
elastic–plastic response provided by the SPME formulation.
Fig. 4a shows the comparison between the experimental response of 6.2. Combined horizontal and centred vertical load (case study No. 2)
the pile group and the one predicted by the SPME for a purely vertical
force directed downwards (case A1 in the dataset by de Sanctis et al., The second case study used for validating the SPME is the Continuum
2021). The macroelement reproduces well the nonlinear response up to Coupled soil-piles-raft model (CC model) recently implemented by
about 75 % of the ultimate vertical load. For larger amplitudes, the ac­ Chanda et al. (2021), illustrated in Fig. 5, that in turn was validated
curacy of the prediction depends on the number of yield surfaces needed against results obtained with small-scale laboratory and full-scale tests.
for capturing the rapid variation of the tangent stiffness: it is observed This study deals with the effect of the simultaneous presence of a vertical
that 10 yield surfaces are sufficient to ensure a smooth transition to and a horizontal load acting on piled foundations in sandy and clayey
failure. While the failure load predicted by the model derives directly soils. The multiaxial response of the foundation was therein investigated
from its calibration against experimental data, failure conditions for by using three-dimensional finite element analyses, modelling explicitly
eccentric loads are a prediction of the model and therefore may not be the interaction of the raft with the surrounding soil. The proposed SPME
identical with the experimental results. This is illustrated in Fig. 4b, was successfully validated against the numerical results obtained in all
where the experimental moment-rotation relationships are compared the configurations considered therein (variability of the soil-raft inter­
with the macroelement calculations for cases A2 and B1 of the experi­ action, of the pile slenderness ratio, of the pile spacing and of the type of
mental dataset, having high and small eccentricity, respectively. connection between piles and raft). For brevity, only the results of some
Although for these cases the SPME provides a slight overestimation of the analyses are shown here, referring to a reference soil-piles-raft system,
failure moment for case A2 (4 %) and a more important underestimation named case no. 2, which is a 3 × 3 pile group with 20 m-long piles
for B1 (19 %), the overall prediction of the loading curves is quite good, hinged to a raft which is in contact with the soil surface. The soil domain
also considering that the model cannot predict any softening, and that the was modelled with solid elements employing the elastic–plastic hard­
ening soil model developed by Schanz et al. (1999), with constitutive

Fig. 5. A) Calculation domain of the coupled soil-piles-raft model of case study no. 2 (reproduced by Chanda et al., 2021).

7
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

parameters calibrated on experimental data (Chanda et al., 2021). It is


important to observe that the piles were represented as embedded beam
elements with a linearly elastic behaviour, and therefore they cannot
yield.
In the CC model a first calculation stage was used to reproduce the
stress state in the soil deposit at the end of the foundation construction.
This was followed by the application of load increments on the raft,
subdivided into two phases: i) application of a constant centred vertical
load, taken as 0, 10 %, 50 %, and 100 % of the ultimate bearing capacity;
ii) progressive increase of the horizontal load Q1.
The SPME is therefore used to simulate the response of the whole
soil-piles-raft system. The plastic domain is completely described by the
specification of the bearing capacities, the horizontal limit force under
uniaxial loading conditions, Q(1D)
1 , and the maximum external moment,
Q(max) (+)
R2 . The vertical bearing capacity Q3 = 75 MN was obtained from
Authors’ numerical results, and accounts for the presence of the
embedded raft (Chanda et al., 2021). By contrast, Q(-) 3 = -13.6 MN was
evaluated through standard design procedures, assuming a null contri­
bution of the raft in tension.
The horizontal limit load of the system was derived by the numerical Fig. 7. Validation of the macroelement response against the results of the
q1-Q1 relationship produced by the Authors as follows. Because of the coupled soil-piles-raft model of Fig. 5 with hinged piles-raft connection: q1-Q1
linear behaviour of the structural members, the numerical results, relationships considering different ratios Q3/Q(+)
3 = 0, 0.1, 0.5.
shown in Fig. 6a for different levels of the mobilised resistance (black
dashed lines), can reasonably simulate the macro-response of the
foundation only for small- to medium-strain levels (far from the soil-raft contact, that increases markedly the limit load for nearly ver­
attainment of the pile strength activating a global plastic mechanism). tical loading, while has a minor effect on the horizontal response of the
Therefore, the expected limit load was evaluated by fitting the numer­ pile group.
ical results with a hyperbolic law, whose coefficients were identified by The remaining ingredient needed to characterise completely the
means of the procedure proposed by Chin (1970, 1971). The resulting monotonic response of the SPME is the initial stiffness matrix. For the
hyperboles (Fig. 6a, grey dashed lines) provide a horizontal limit load present case, H(0) (0)
11 and H33 are the only components required for the
varying from 9.8 MN to 11.4 MN for Q3 / Q(+) 3 = 0.0 and 0.5, respec­ initial stiffness matrix, that were estimated directly from the numerical
tively, computed as 95 % the respective horizontal asymptotes. results (Chanda et al., 2021) as 44.3 MN/m and 232.1 MN/m,
The resulting plastic domain of the SPME is shown in Fig. 6b as 8 respectively.
yield surfaces in the Q1-Q3 space. In this case, these loci show a quasi- Fig. 7 shows the comparison between the q1-Q1 responses obtained
elliptical shape, different from the marked ovoidal shape obtained for with the Coupled Continuum (CC) model of the reference soil-piles-raft
the case of Fig. 2. This difference is due to the stress exchanged at the

Fig. 6. A) Determination of the horizontal limit loads for ratios Q3/Q(+)


3 = 0, 0.1, 0.5 and b) representation of the plastic domain of the macroelement in the Q1-
Q3 space.

8
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

system and with the macroelement (SPME), relative to the three additional moment produced by the horizontal force with respect to the
different loading paths illustrated in Fig. 6b. The comparison is master point (vertical distance between the raft extrados and the master
restricted to the deformation interval investigated with the CC model. point equal to 0.5 m).
The response of the piled foundation is visibly nonlinear, starting from The calibration of the macroelement was based on the evaluation of
small levels of deformation. This is reproduced quite well by the the ultimate capacity and the initial stiffness under multiaxial condi­
piecewise linear response of the SPME for all the loading ratios Q3 / Q(+)
3 tions. The hyper-ovoidal ultimate yield surface was calibrated on the
= 0, 0.1 and 0.5. The response becomes stiffer as the vertical load rises, sole information about the bearing capacity of the group, for Q1/Q3
and this result is well reproduced by the SPME: in fact, the initial stiff­ equal to 0 and to 0.577, and by using the non-dimensional abaci by
ness for Q3 = 0 was used as an input quantity in the calibration, while Gorini and Callisto, 2022a) to account for the effect of the external
the stiffness of the other curves is a prediction of the macroelement, moment on the combined Q1-Q3 failure load. The bearing capacity for
produced as a combination of the elastic–plastic stiffnesses H11 and H33. Q1/Q3 equal to 0 and to 0.577 was obtained by the force–displacement
In these cases, the SPME reproduces well both the initial stiffness and its curves provided by Franza and Sheil (2021): it was taken as the load
evolution with the level of mobilised strength, validating the hyperbolic corresponding to a tangent stiffness equal to 10 % of the initial one (see
variation of the stiffness taken as a basis for the kinematic hardening law failure points in Fig. 8a and 9a). The plastic domain was then defined
of the model (Gorini et al., 2022a). accordingly, considering 8 yield surfaces. The initial stiffness tensor of
the SPME was obtained by using the solution for the elastic stiffness of
embedded foundations provided by Gazetas (1991).
6.3. Case study no. 3: pile group under horizontal and inclined vertical The comparison between the two approaches refers to load paths
load characterised by an eccentric vertical load and a horizontal one, the
latter producing a moment having the same direction as the one pro­
The response of the SPME to the concomitant presence of an external duced by the vertical load. Two combinations Q1/Q3 = 0, 0.58 are taken
moment and of an inclined force is tested with reference to the case into consideration for testing the SPME response from small to large
study published by Franza and Sheil (2021). In this study, a group of strain levels.
identical 3 × 3 piles is considered, that are fully connected to a rigid Fig. 8 shows the foundation response in the case of an eccentric
suspended raft (no soil-raft interaction) and embedded in homogeneous vertical load. The vertical response, represented in the normalised space
undrained clayey soil. The piles are characterised by a diameter D = 0.5 of Fig. 8a, is reproduced reasonably well by the SPME considering the
m, an embedded length L = 12.5 m and a spacing s = 5D. A Continuum large variation of the normalised eccentricity ex/D = 0–10. The SPME
Coupled (CC) numerical representation of the soil-foundation system attains the ultimate capacity in proximity of the estimated failure loads
was implemented by Franza and Sheil (2021) using Plaxis 3D. The soil of the CC model. The SPME is also able to simulate the rotational
domain was therein regarded as a weightless medium exhibiting an response of the foundation (Fig. 8b), providing a modest underestima­
elastic-perfectly plastic behaviour combined with the Tresca failure tion of the ultimate capacity.
criterion. The reinforced concrete piles were modelled as linear elastic The case of a triaxial load Q1-Q3-QR2 is illustrated in Fig. 9 (Q1 =
solid elements. 0.58Q3). The two models are qualitatively in a good agreement. The
In Franza and Sheil (2021), the resultant horizontal-vertical force is presence of the horizontal force reduces the limit vertical load and,
applied to the extrados of the raft at a distance ex (eccentricity) from the therefore, the rotational capacity of the foundation (cf. Figs. 8 and 9).
vertical central axis of the foundation, whereas the foundation dis­ The response is highly nonlinear starting from small strain levels and the
placements refer to the central point on the ground level (master point). SPME produces deformations that are dominated by the plastic regime.
The generalised force–displacement relationship of the SPME was This feature represents the macro-scale effect of the irreversible
therefore assumed to be lumped at the master point by considering the

ex
ex
ex
ex
Fig. 8. Comparison of the macroelement response with the results of fully coupled numerical pushover analyses (Case 3) under an eccentric vertical load: a) vertical
and b) rotational response considering different eccentricity ratios e/D.

9
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

Fig. 9. Comparison of the macroelement response with the results of fully coupled numerical pushover analyses (Case 3) under an inclined and eccentric load (Q1 =
0.58Q3): a) vertical and b) rotational response considering different eccentricity ratios e/D.

behaviour of both soil and piles, as the global plastic mechanism for Q1 The SPME calibration was devoted to the identification of the ulti­
= 0.58Q3 is expected to involve the flexural capacity in the piles (Gorini mate yield surface and of the foundation response at small-deformation
and Callisto, 2022a; Sakellariadis and Anastasopoulos, 2022). The flow levels. The ultimate surface derives from the definition of the general­
rule of the SPME simulates properly the direction of the generalised ised failure loads along specific load paths, that can be straightforwardly
plastic deformations. The differences become more evident when the evaluated through standardised procedures, and of the three parameters
vertical load is not centred, and in this case the SPME provides a more Sf0, aG and aQ3 whose optimum values were identified as described in
rapid decrease of the generalised failure load as ex rises, if compared to Section 5.1. The assigned input parameters are collected in Table 1, and
the CC model. In these conditions, the response of the CC model never the corresponding plastic domain is shown in Fig. 10c, only in the Q1-Q3
reaches a vertical asymptote and predicts a large rotational capacity. space for brevity, considering 5 yield surfaces.
This feature can be ascribed to the linear behaviour assumed for the The small-strain frequency-dependent response of the SPME derives
structural members in the CC model. from the combination of its initial stiffness, H(0)ij , with the participating
masses of the soil-piles-raft system. Their identification was obtained
7. Dynamic Response of the Macroelement from impedance functions computed with the software DYNA (http
s://www.eng.uwo.ca/grc/dyna6/index.html). As an example, Fig. 11
7.1. Case study no. 4 and dynamic identification shows the variation of the longitudinal dynamic stiffness of the soil-piles
system H(DYN)
11 with the vibration period T (the raft was assumed with no
The dynamic features of the SPME are now studied with reference to mass at this stage). From this relationship the stiffness component H(0) 11
a piled foundation recently analysed in a wide research project on the = H(1) (st) (0)
11 = H11 and the fundamental vibration period T1 were deter­
actual seismic risk of bridges designed according to the Italian technical mined as shown in the figure. The corresponding participating mass was
provisions (ReLUIS-DPC 2019–2021). In that study several structural evaluated accordingly as m1 = 8π2 H(0) 11 × T1
(0)2
+ m(raft), where m(raft) is
layouts were analysed and the present paper focusses on the foundation the mass of the raft. The resulting small-strain, fundamental vibration
of pier P2 of a tall viaduct located in L’Aquila, central Italy (see Gorini periods of the foundation system are equal to 0.09 s, 0.09 s, 0.08 s, 0.05 s
and Callisto, 2022b, for a more detailed description of the case study). and 0.05 s for the degrees of freedom 1, 2, 3, R1 and R2, respectively.
Pier P2 is supported by the piled foundation schematically illustrated in
Fig. 10a: the foundation is composed of 4 × 7 reinforced concrete piles
7.2. Frequency-dependent response at small deformation levels
having a length of 54 m and the dimensions in plan shown in the figure.
The subsoil is composed of silty and clayey layers, with an equivalent
The ability of the SPME to simulate the essential characters of the
shear wave velocity in the top 30 m of depth in correspondence of pier
dynamic response of piled foundations is investigated through a series of
P2 equal to vs30 = 650 m/s and bedrock, vs = 1000 m/s, located at a
force-controlled dynamic analyses carried out in OpenSees. The finite
depth of 63 m from the foundation level (pile tips are 8 m above the
element representing the SPME was perturbed by a constant amplitude,
bedrock). The strength of the superficial soil layers is described by a
harmonic external force, Qi, for ten cycles in correspondence of the node
cohesion of 10 kPa and angle of shearing resistance in the range
which the participating mass is applied to, keeping the other node fixed.
24◦ –26◦ . The piles were designed under static and seismic conditions
As a first step, the amplitude of Qi was chosen to activate the first plastic
(Italian Building Code, IBC, 2018), referring to the seismic demand of
flow only, leading to a nearly linear response associated with the small-
the Apennine area of central Italy. The soil-raft interaction was neglec­
displacement vibration periods for which the SPME was calibrated. In
ted as a common assumption in standard practice. The resulting
this parametric study, the ratio of the period T of Qi to the fundamental
moment-axial force strength envelope of the pile cross-section is
period T(0)
i ranges between 0.05 and 22; for each vibration period, the
depicted in Fig. 10b.
maximum deformation of the SPME, q(max) i , was computed.

10
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

Fig. 10. A) Plan view of the piled foundation tested under dynamic conditions, b) bending moment-axial force strength envelope of the pile cross-section and c)
plastic domain of the associated macroelement in the Q1-Q3 space.

Fig. 12 shows the relationships between the period T and the corresponding to a period elongation of 67 %. The progressive shift in
displacement dynamic amplification ratio q(max)i / q(st)
i (q(st)
i is the SPME the dominant period is accompanied by a modest reduction of the dy­
(0)
deformation for T / Ti = 22). Of course, the maximum amplification namic amplification. This stabilisation of the dominant period and of the
occurs at the fundamental periods found in the previous section. When a maximum amplification simulates the steady-state dynamic response of
single force component is applied dynamically, the maximum amplifi­ geotechnical systems associated with the activation of a global plastic
cation is obtained in the vertical direction, as a consequence of the larger mechanism (Gorini, 2019: Gorini et al., 2022a).
mass involved in the vertical response of the system. However, a com­ Because of the presence of a participating soil mass, also the energy
bined dynamic loading, in which both the vertical and horizontal force dissipation of the macroelement response becomes influenced by the
components are applied at the same time with no phase difference, is period of the external perturbation. This is evident in Fig. 14, showing
seen to produce the maximum amplification, as shown in Fig. 12d, the internal force–deformation cycles of the SPME in the longitudinal
where q(max)
13 = max{(q21 + q23)0.5}. direction, corresponding to a high level of the mobilised strength (nact =
4–5). When T / T(steady)
1 = 1 (T(steady)
1 = longitudinal vibration period for
7.3. Effect of the nonlinear behaviour of the soil-foundation system nact = 4–5), the cycles open due to the magnified inertial effects in the
SPME. The internal force reaches the ultimate yield surface, associated
In the previous section it was seen that SPME can reproduce the with a limit value of 26 MN. For sufficiently higher periods (T / T(steady)
1
multiaxial dynamic features of piled foundations at small-displacement > 10) the inertial effects become negligible and energy dissipation is
levels. However, the participating masses of the SPME interact also with basically relative to a quasi-static application of the loads, while for very
the activated plastic flows in the case of more severe external pertur­ low normalised periods (T / T(steady)
1 ≤ 0.2) the macroelement response
bations, and this modifies the dynamic response of the system, attenuates significantly due to its out-of-phase motion compared to the
enhancing the plastic response of the soil. oscillations of the load. Nonetheless, in the considered range of the
Fig. 13 shows the dynamic amplification curves of the SPME in di­ mobilised longitudinal resistance Q1/ Q(lim)
1 = 0.7–1.0 the damping ratio
rection 1 obtained by repeating the dynamic analyses in Section 7.2 with is not affected by the loading frequency, being of about 31 % for all the
larger amplitudes of the longitudinal load, obtaining the activation of an cycles shown in Fig. 14.
increasing number of plastic flows, nact, from 1 to N = 5. The funda­
mental period increases at the attainment of the second yield surface,
and then stabilises from the attainment of the third plastic flow on,

11
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

Table 1
Calibration of the input parameters of the macroelement for case no. 4.
Variable Units Description value

ultimate yield surface Q(1D)


1 = Q(1D)
2 MN horizontal limit loads of the pile group 26.4
Q(+)
3 MN vertical limit load of the pile group in compression 245.5
(-)
Q3 MN vertical limit load of the pile group in tension − 180.1
Q(max)
R1 MNm maximum external moment around axis 1 1232.2
Q(max)
R2 MNm maximum external moment around axis 2 711.8
S(0) (0)
f,1 = Sf2 – size parameter of the ultimate surface along the Q1-2-axis for QR2-R1 = 0 0.918
aΓ,1 = aΓ,2(0) – size parameter of the ultimate surface along the Q1-2-axis for |QR2-R1|>0 0.8
aQ3,1 = aQ3,2 MN shape parameter of the ultimate surface in the Q1-Q3 space 421.0
N - number of yield surfaces 5
initial stiffness H(0)
11 MN/m initial stiffness of the soil-piles system in direction 1 1.72 × 104
H(0)
22 MN/m initial stiffness of the soil-piles system in direction 2 1.63 × 104
H(0)
33 MN/m initial stiffness of the soil-piles system in direction 3 2.63 × 104
H(0)
R1R1 MNm initial rotational stiffness of the soil-piles system around axis 1 4.75 × 106
H(0)
R2R2 MNm initial rotational stiffness of the soil-piles system around axis 2 2.32 × 106
modal masses m11 Mg participating mass of the soil-piles system in direction 1 1.76 × 103
m22 Mg participating mass of the soil-piles system in direction 2 1.74 × 103
m33 Mg participating mass of the soil-piles system in direction 3 2.29 × 103
mR1R1 Mgm2 participating rotational inertia of the soil-piles system around axis 1 1.65 × 105
mR2R2 Mgm2 participating rotational inertia of the soil-piles system around axis 2 8.34 × 104

identified by using the procedure in Section 5. The resulting small-strain


dynamic responses of the geotechnical systems are characterised by
fundamental vibration periods of 0.15 s to 0.25 s for the abutments and
of 0.05 s to 0.11 s for the piers foundations. The piers were modelled as
displacement-based beam elements reproducing the properties of the
effective reinforced concrete cross section. The Kent-Scott-Park model
(Scott et al., 1982) was assigned to the concrete fibers and an elastic–­
plastic material with kinematic hardening to the steel fibers. The stiff­
ness and strength of the bearing devices were simulated through the
combination of nonlinear rheological elements, while the deck was
reproduced through equivalent elastic beam elements.
In the global structural model with TIMs, a staged analysis procedure
was adopted, composed of a gravity analysis and the subsequent appli­
cation of the seismic motion. The latter is represented by the three-
component displacement time histories of a near-collapse earthquake
scenario for the bridge at hand (Gorini and Callisto, 2022b). The per­
formance of the TIMs is shown in Fig. 16 in terms of the generalised
Fig. 11. Translational dynamic stiffness in direction 1, H(dyn)
internal force–deformation cycles along the translational and rotational
11 , of case no. 4
plotted as a function of the period. degrees of freedom. The marked nonlinear behaviour of the soil-piles
systems causes significant permanent effects in the geotechnical sys­
tems that alter the bridge performance and the displacement demand for
7.4. Example application: seismic assessment of a multi-span tall viaduct
the bearings. The abutments show a pronounced nonlinear response
because of the embankment mass participating to the bridge response.
As an application of the proposed TIM approach in a dynamic
The abutments attain the active resistance and accumulate irreversible
analysis of a structure, the TIMs for abutments (Gorini et al., 2023) and
displacements inwards. The piers foundations are more rigid than the
piled foundations are now employed to analyse the seismic performance
soil-abutment systems, with a frequency response that, at least in the
of the case study shown in Fig. 15, where the foundation of pier P2
horizontal directions, is marginally coupled with the dynamic response
represents the case study introduced in Section 7.1. The main characters
of the viaduct (first vibration period of the superstructure in the longi­
of the case study are provided in the following, while the reader can
tudinal direction equal to about 1 s). The consequent, more limited
refer to Gorini and Callisto (2022b) for a more comprehensive descrip­
dynamic amplification of the foundation response produces lower,
tion of the soil-bridge layout and of the relative mechanical properties.
though not negligible, irreversible displacements, with a maximum of
The bridge rests on five supports and crosses a V-shaped valley with
0.02 m at pier P2. Nonetheless, the nonlinear behaviour produces the
stiffness and strength properties of the layered subsoil discussed in
development of remarkable bending moment and rotations at the piers
Section 7.1.
base (Fig. 16k-m), with significant permanent values in correspondence
According to the static scheme of the bridge, in the longitudinal di­
of the piers presenting an integral connection with the deck, P2 and P3.
rection the deck is connected only to piers P2 and P3. The abutment A2
Finally, the analysis a posteriori of the inflection of the piles within
is connected longitudinally to the deck using viscous dampers, while the
the active length provided information about damage in the structural
relative deck-abutment displacement is free at the abutment A1. In the
members. The maximum horizontal deformations q1,max of the SPMEs
transverse direction, the deck is connected to all supports. The diameter
were compared with the respective thresholds, q1,y, corresponding to
of the foundation piles ranges between 1.0 and 1.2 m.
yielding of the piles. For each pier foundation, the former value was
The seismic performance of the bridge was investigated using a nu­
obtained by the dynamic analysis of the bridge with macroelements,
merical representation of the bridge model with the TIMs implemented
whereas the yield threshold was computed through the following steps:
in OpenSees. The calibration of the macroelements for the abutments,
i) evaluation of the yield moment My of the pile cross section, referring
here omitted for brevity, follows the procedure in Gorini et al. (2023),
to the axial force under static conditions; ii) use of the Winkler elastic
while the properties of the SPME for the piers foundations were
solution for long, fixed-head piles (Viggiani et al., 2012) to evaluate the

12
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

Fig. 12. Dynamic amplification curves at small deformations (q(max)


i is the maximum deformation of the SPME in the i-th degree of freedom): a,b) translations along
directions 1 and 3, c) rotation around axis 1 and d) combined response in the horizontal-vertical force space.

Fig. 13. Dynamic amplification curves of the SPME in direction 1 from low to Fig. 14. Force-deformation cycles of the SPME in direction 1 at high levels of
high levels of the mobilised strength of the soil-piles system (nact = number of the mobilised strength for different values of the normalised period T/T(steady)
1

plastic flows activated during the analysis). = 0.2, 1.0, 5.0, 13.0 (T(steady)
1 = 0.15 s is the SPME fundamental vibration period
for high levels of the mobilised strength).

13
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

Fig. 15. Longitudinal section of the reference soil-bridge system (lengths in meters).

A1 P1 P2 P3 A2
50 (a) (b) (c) (d) (e)
long. force Q1 (MN)

active side
25 active side
0

-25

-50
0 0.1 0.2 -0.03 0 0.03 -0.03 0 0.03 -0.03 0 0.03 0.2 0.1 0
q1 (m) q1 (m) q1 (m) q1 (m) q1 (m)
transv. force Q2 (MN)

50 (f) (g) (h) (i) (j)


post-earthquake
25 deformed shape

-25

-50
0 0.1 0.2 -0.03 0 0.03 -0.03 0 0.03 -0.03 0 0.03 0.2 0.1 0
transv. disp. q2 (m) q2 (m) q2 (m) q2 (m) q2 (m)

300 (k) (l)


moment QR2 (MNm)

150

-150
(m)
-300
-0.0004 0 0.0004 -0.0004 0 0.0004 -0.0004 0 0.0004
rotation qR2 (rad) qR2 (rad) qR2 (rad)

Fig. 16. Seismic performance of the geotechnical systems of the reference bridge in terms of a-e) longitudinal, f-j) transverse and k-m) rotational (about axis 2)
force–deformation responses of the macroelements, with post-earthquake values shown by the filled circles.

horizontal inflection of the piles q1,y that produces a bending moment 8. Conclusions
equal to My. In this computation, the soil-pile relative stiffness was
chosen to provide a relationship q1,y-Q1 consistent with the constitutive The conventional approaches to the design of pile groups, based on a
relationship provided by the SPME. Using this approach, it was separate assessment of the effects of the vertical and horizontal loads,
demonstrated that all piles yielded during the considered seismic event, rely on considerable experience and have the merit of being endowed
highlighting the importance of accounting for such peculiar features of with a clear physical meaning for simple loading conditions. However,
the response in the assessment of existing bridges. the use of these approaches becomes problematic for complex loading

14
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

conditions, and particularly for the highly variable and multi-directional CRediT authorship contribution statement
foundation loads produced by the seismic action. This is particularly
evident for the seismic design of important bridges, which is becoming Davide Noè Gorini: Conceptualization, Methodology, Software,
more and more based on time-domain nonlinear dynamic analyses, in Validation, Formal analysis, Investigation, Data curation, Writing –
which the nonlinear behaviour of the foundation concurs to the overall original draft, Writing – review & editing, Visualization. Luigi Callisto:
seismic performance of the bridge. These exigences call for an analysis Conceptualization, Methodology, Validation, Writing – review &
approach capable to consider complex features, like the directional editing.
coupling, the dynamic amplification, the nonlinear and cyclic response,
the instantaneous and permanent deformation pattern. On the other Declaration of Competing Interest
hand, it is auspicable that the experience gained in the traditional
methods for the evaluation of the one-directional and monotonic The authors declare that they have no known competing financial
behaviour of pile groups could be incorporated in the advanced models, interests or personal relationships that could have appeared to influence
to contribute to their soundness. the work reported in this paper.
In fact, the formulation of the macroelement proposed in the present
paper, based on a rigorous thermodynamic approach and developed Data availability
within the context of hardening plasticity, is conceived to be calibrated
on the basis of such traditional methods: the ultimate surface can be Data will be made available on request.
found on the basis of any bearing capacity solution for vertical and
horizontal loads, separately, and the calibration of the small-strain Acknowledgements
response can be obtained from widely available elastic solutions. This
ease of calibration, together with the satisfactory response and the very Part of this research was sponsored within the activities of the
limited additional computational demand associated with its use, make ReLUIS-DPC and DPC 2019–2021 research programs, funded by the
the macroelement a very powerful tool for the nonlinear analysis of Presidenza del Consiglio dei Ministri—Dipartimento della Protezione
structures accounting for the effect of soil-structure interaction. The Civile (DPC).. The Authors acknowledge the contribution of Prof. P.
implementation of the model into a zero-length finite element in Franchin and Dr. F. Noto for the development of the structural model of
OpenSees makes the use of this model widely available and will hope­ Fig. 15.
fully contribute to its extensive use within the scientific and technical
community.

Appendix:. Details on the Incremental Response

Gradient of the nth yield surface

∂y(n)
The derivative in Eq. (15) encloses the following composed functions:
∂χ(n)
3

( ) 2
( ) ( ) ( ( ) ( ) )2
̂ 3 χ (n) • ̂
F1 = A Ri S F,1 χ (n) ̂ 2 (n) c 3 χ (n)
R2 • Γ 1 χ R2 • ̂ Ri
̂ 3,1 χ (n)
+Q R2 (21)

( ) 2
( ) ( ) ( ( ) ( ) )2
̂ 3 χ (n)
F2 = A Ri S F,2 χ (n)
•̂ ̂ 2 (n) c 3 χ (n)
R1 • Γ 2 χ R1 • ̂ Ri
̂ 3,2 χ (n)
+Q R1 (22)

( ( ) )2
(23)
(n) (n) (n) (n) ̂ 3,1 χ(n)
yN1 = 2 • χ1 • χ3 + χ1 • Q R2

( ( ) )2
y(n) (n) (n) (n) ̂ (n)
N2 = 2 • χ2 • χ3 + χ2 • Q 3,2 χR1 (24)
( )
̂ 3 χ(n)
A R2
( )
y(n)
D1 = − χ (n)
3 + c 3 χ(n)
+̂ R2 (25)
2
( )
̂ 3 χ(n)
A R2
( )
y(n)
D2 =χ (n)
3 + − ̂c 3 χ(n)
R2 (26)
2
( )
̂ 3 χ(n)
A R1
( )
y(n)
D3 = − χ (n)
3 + + ̂c 3 χ(n)
R1 (27)
2
( )
̂ 3 χ(n)
A R1
( )
y(n)
D4 =χ (n)
3 + − ̂c 3 χ(n)
R1 (28)
2
( )
Whereas in the derivative ∂y(n) /∂χ(n)
R1 (Eq.(16)), function yD5 χR1
(n) (n)
is so defined:
( ̂3 ) ( ̂3 )
2 ( 2 ) A A
(n) ̂3 • ̂
yD5 = A ̂2 • ̂
S F,2 • Γ ̂ 3,2 2 •
c3 + Q (n)
− χ3 +
(n)
+ ̂c 3,2 • χ3 + ̂3 • ̂
c 3,2 = A
− ̂
2
̂ 2 • yD5,1 • yD5,2 • yD5,3
S F,2 • Γ (29)
2 2
2 2
( 2 )
̂ 3,2 2
c3 + Q
yD5,1 = ̂ (30)

15
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

̂3
A
yD5,2 = − χ(n)
3 + c 3,2
+̂ (31)
2

̂3
A
(32)
(n)
yD5,3 = χ3 + − ̂c 3,2
2
The first addendum of ∂y(n) /∂χ(n)
R1 develops as:

∂y(n) ∂Q3,2 ( )
N2
(n)
= (n) • 4 • χ(n) (n) (n) (n) ̂
2 • χ2 • χ3 + χ2 • Q 3,2 (33)
∂χR1 ∂χR1

where:
∂Q3,2
nQ3,2
aQ3,2 ⃒ ⃒[nQ3,2 − 1] ( )
⃒ ⃒ (1− nQ3,2 )
(n)
= − ( )nQ3,2 • ⃒χ(n)
R1 ⃒ • sgn χ(n)
R1 • Q3,2 (34)
∂χR1 (max)
QR1 • bQ3,2

Hence, substituting Eq. (34) into Eq.(33), the latter becomes:

∂ y(n)
nQ3,2
aQ3,2 ⃒ ⃒[nQ3,2 − 1] ( ) ( )
⃒ ⃒ (1− nQ3,2 )
)nQ3,2 • ⃒χ(n) (35)
N2 (n) (n) (n) (n) (n) ̂ 3,2
(n)
=− ( R1 ⃒ • sgn χR1 • Q3,2 • 4 • χ2 • χ2 • χ3 + χ2 • Q
∂χ R1
(max)
QR1 • bQ3,2

The second part of Eq. (16) develops as follows:


⎡ ( ) ⎤
∂y(n) ∂ ( ̂ ̂2 2
) ∑ 6 ∂Gi,2 χ(n)
R1
( )
(36)
D5 ̂ ⎣ (n) ⎦
(n)
= (n) A 3 • S F,2 • Γ 2 • YD5,1 • YD5,2 • YD5,3 = (n) =i Gj,2 χR1
• Πj∕
∂χR1 ∂χR1 i=1 ∂χR1

( )
where Gj,2 χR1 is the generic function in Eq. (29) depending on the nth dissipative moment χR1 . The derivatives in Eq. (36) are developed below:
(n) (n)

̂ 3 /∂χ (n)
• ∂A R1

̂ 3,2
∂A
nA3,2
aA3,2 ⃒ ⃒(nA3,2 − 1) ( )
= − (

)nA3,2 • ⃒χ(n)
R1 ⃒

• sgn χ(n)
R1
̂ (1−
•A
nA3,2 )
(37)
(n) 3,2
∂χR1 (max)
QR1 • bA3,2

2
S F,2 /∂χR1
(n)
• ∂̂
2
∂ ̂S F,2
nSf ,2
aSf,2 ⃒ ⃒(nSf ,2 − 1) ( )
⃒ ⃒ (2− nSf ,2 )
)nSf ,2 • ⃒χ(n) (38)
(n)
=− 2•( R1 ⃒ • sgn χR1 • ̂S F,2
∂χ(n)
R1
(max)
QR1 • bSf,2

̂ 2 /∂χ(n)
• ∂Γ 2 R1

̂2
∂Γ
nΓ,2
aΓ,2 ⃒

⃒(nΓ,2 − 1)

( ) (2− nΓ,2 )
)nΓ,2 • ⃒χ(n) (39)
(n)
2
(n)
= − 2 • ( R1 ⃒ • sgn χR1 • ̂S Γ,2
∂χR1 (max)
QR1 • bΓ,2

• ∂yD5,1 /∂χR1
(n) (n)

∂y(n) √̅̅̅̅̅̅̅̅̅ ( ̂ 3,2


)
∂̂c 3,2 ∂ Q
(40)
D5,1
=2• y(n)
D5,1 • +
∂χ(n)
R1
(n)
∂χR1(n)
∂χR1

∂̂c 3,2
nC3,2
ac3,2 ⃒ ⃒(nC3,2 − 1) ( )
⃒ ⃒ (1− nC3,2 )
(n)
=− ( )nC3,2 • ⃒χ(n)
R1 ⃒ • sgn χ(n)
R1 c 3,2
•̂ (41)
∂χR1 (max)
QR1 • bc3,2

so Eq. (40) becomes:

16
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

⎡ ⎤
∂y(n) ( )
nC3,2
ac3,2 ⃒ ⃒(nC3,2 − 1) ( ) nQ3,2
ac3,2 ⃒ ⃒(nQ3,2 − 1) ( )
D5,1 ̂ 3,2 • ⎢ ⃒ ⃒ (1− nC3,2 ) ⃒ ⃒ (1− nQ3,2 ) ⎥
̂ 3,2 + Q
=− 2• C ⎣( )nc3,2 • ⃒χ(n)
R1 ⃒ • sgn χ(n)
R1 • ̂c 3,2 +( )nQ3,2 • ⃒χ(n)
R1 ⃒ • sgn χ(n)
R1 • Q3,2 ⎦
∂χ(n)
R1
(max)
QR1 • bc3,2 Q(max)
R1 • bQ3,2
(42)

• ∂yD5,2 /∂χR1
(n) (n)

∂y(n) ̂ 3,2 ∂ C
∂A ̂ 3,2 nA3,2
aA3,2 ⃒ ⃒(nA3,2 − 1) ( ) nc3,2
ac3,2 ⃒ ⃒(nc3,2 − 1) ( )
D5,2
= 0.5 • + = − 0.5 • (

)nA3,2 • ⃒χ(n)
R1 ⃒

• sgn χ(n)
R1
̂ (1−
•A
nA3,2 )
− (

)nc3,2 • ⃒χ(n)
R1 ⃒

• sgn χ(n)
R1
(1−
c 3,2
•̂
nc3,2 )
(n) (n) (n) 3,2
∂χ R1 ∂χR1 ∂χR1 (max)
QR1 • bA3,2 (max)
QR1 • bc3,2
(43)

• ∂yD5,3 /∂χR1
(n) (n)

∂y(n) ̂ 3,2 ∂̂c 3,2


∂A
nA3,2
aA3,2 ⃒ ⃒(nA3,2 − 1) ( ) nc3,2
ac3,2 ⃒ ⃒(nc3,2 − 1) ( )
D5,3
= 0.5 • − = − 0.5 • (

)nA3,2 • ⃒χ(n)
R1 ⃒

• sgn χ(n)
R1
̂ (1−
•A
nA3,2 )
+(

)nc3,2 • ⃒χ(n)
R1 ⃒

• sgn χ(n)
R1
(1−
• ̂c 3,2
nc3,2 )
3,2
∂χ(n)
R1
(n)
∂χR1 ∂χR1 (n) (max)
QR1 • bA3,2 (max)
QR1 • bc3,2
(44)

Finally, ∂y /∂χ is obtained by introducing Eqs. (35)-(44) into Eq.(16).


(n) (n)
R1

In the space χ1 -χ3 -χR2 , the yield surfaces are still represented by an ovoidal surface with coefficients depending on the dissipative moment χR2 . It
(n) (n) (n) (n)

follows that the derivative ∂y(n) /∂χ has the same analytical form as Eq.(16), in which however Eqs. (35)-(44) consider the evolution functions ̂I i (QR2 )
(n)
R2
(Eqs. (3)-(7)) depending on χR2 .
(n)

References Gorini, D.N., 2019. Soil-structure interaction for bridge abutments: two complementary
macro-elements. PhD thesis, Sapienza University of Rome, Italy, https://iris.uni
roma1.it/handle/11573/1260972.
Boulanger, R.W., Curras, C.J., Kutter, B.L., Wilson, D.W., Abghari, A., 1990. Seismic soil-
Gorini, D.N., Callisto, L., 2022a. Generalised ultimate loads for piled foundations. Acta
pile-strcture interaction experiments and analysis. J. Geotech. Geoenviron. Eng.
Geotech. 17, 2495–2516. https://doi.org/10.1007/s11440-021-01386-4.
ASCE 125 (9), 750–759.
Gorini, D.N., Callisto, L., 2022b. A class of thermodynamic inertial macroelements for
Chanda, D., Nath, U., Saha, R., Haldar, S., 2021. Development of lateralcapacity-based
soil-structure interaction. In: Springer Series in Geotechnical, Geological and Earthquake
envelopes of piled raft foundation under combined V-M–H loading. Int. J. Geomech.
Engineering. Design inEarthquake Geotechnical Engineering, Beijing, pp. 1095–1102.
21 (6), 04021075. https://doi.org/10.1061/(ASCE)GM.1943-5622.0002023.
https://doi.org/10.1007/978-3-031-11898-2_87.
Chatzigogos, C.T., Figini, R., Pecker, A., Salencon, J., 2011. A macroelement formulation
Gorini, D.N., Callisto, L., 2020. A macro-element approach to analyse bridge abutments
for shallow foundations on cohesive and frictional soils. Int. J. Numer. Anal. Meth.
accounting for the dynamic behaviour of the superstructure. Geotechnique 70 (8),
Geomech. 35 (8), 902–931.
711–719. https://doi.org/10.1680/jgeot.19.ti.012.
Chin, F.K., 1971. Discussion on pile test. ASCE Journal for Soil Mechanics and
Gorini, D.N., Callisto, L., Whittle, A.J., 2022a. An inertial macroelement for bridge
Foundation Engineering 97, 930–932.
abutments. Geotechnique 72 (3), 247–259. https://doi.org/10.1680/jgeot.19.P.397.
Chin, F.K., 1970. Estimation of the ultimate load of piles not carried to failure. Proc. 2nd
Gorini, D.N., Callisto, L., Whittle, A.J., Sessa, S., 2022. A multi-axial macroelement for
S.E.Asia Conf. on Soil Eng., pp. 81–90.
bridge abutments. Int. J. Numer. Anal. Meth. Geomech.
Collins, I.F., Houlsby, G.T., 1997. Application of thermomechanical principles to the
Houlsby, G.T., Abadie, C., Beuckelaers, W., Byrne, B., 2017. A model for nonlinear
modeling of geotechnical materials. Proc. R. Soc. Lond. A 453, 1975–2001.
hysteretic and ratcheting behaviour. Int. J. Solids Struct. 120, 67–80.
Correia, A.A., 2011. A Pile-Head Macro-Element Approach to Seismic Design of
Houlsby, G.T., Puzrin, A.M., 2006. Principles of hyperplasticity. Springer, Berlin.
Monoshaft-Supported Bridges. PhD thesis: ROSE School Università degli Studi di Pavia
Italian Building Code, 2018. Norme Tecniche per le Costruzioni. D.M. 14.01.2008, Italian
& Istituto Universitario di Studi Superiori, Pavia, Italy.
Ministry of Infrastructures and Transportation, Rome (in Italian).
Cremer, C., Pecker, A., Davenne, L., 2001. Cyclic macro-element for soil-structure
Karatzia, X., Mylonakis, G., 2012. Horizontal response of piles in layered soil: Simple
interaction: material and geometrical non-linearities. Int. J. Numer. Anal. Meth.
analysis. 2nd International Conference on Performance-Based Design in Geotechnical
Geomech. 25 (13), 1257–1284.
Engineering, Taormina, Italy.
Cremer, C., Pecker, A., Davenne, L., 2002. Modelling of nonlinear dynamic behaviour of
Kaynia, A.M., Kausel, E., 1982. Dynamic Stiffness and Seismic Response of Pile Groups.
a shallow strip foundation with macro-element. J. Earthq. Engng 6 (2), 175–211.
Massachusetts Institute of Technology, Cambridge, Massachusetts. Research Report
de Sanctis, L., Di Laora, R., Garala, T.K., Madabhushi, S.P.G., Viggiani, G.M.B.,
R82–03.
Fargnoli, P., 2021. Centrifuge modelling of the behaviour of pile groups under
Le Pape, Y., Sieffert, J.P., 2001. Application of thermodynamics to the global modelling
vertical eccentric load. Soils Found. 61 (2), 465–479.
of shallow foundations on frictional material. Int. J. Numer. Anal. Meth. Geomech.
Di Laora, R., de Sanctis, L., Aversa, S., 2019. Bearing capacity of pile groups under
25 (14), 1377–1408.
vertical eccentric load. Acta Geotech. 14 (1), 193–205.
McKenna, F., 1997. Object-oriented finite element analysis: Frameworks for analysis,
Di Prisco, C., Nova, R., Sibilia, A., 2003. Shallow footing under cyclic loading:
algorithms and parallel computing. Univ. of California, Berkeley, CA. Ph.D.
experimental behaviour and constitutive modelling. In: Maugeri, M., Nova, R. (Eds.),
dissertation,.
Geotechnical analysis of the seismic vulnerability of historical monuments. Patron,
McKenna, F., Fenves, G.L., Scott, M.H., Jeremic, B., 2000. Open system for earthquake
Bologna, pp. 99–122.
engineering simulation, http://opensees.berkeley.edu.
Dobry, R., Vicente, E., O’Rourke, M.J., Roesset, J.M., 1982. Horizontal stiffness and
Mylonakis, G., Roumbas, D., 2001. Dynamic stiffness and damping of piles in
damping of single piles. J. Geotech. Eng. Div. 108 (3), 439–459.
inhomogeneous soil media. Proceedings, 4th International Conference on Recent
El Naggar, M.H., Bentley, K.J., 2000. Dynamic analysis for laterally loaded piles and
Advances in Geotechnical Earthquake Engineering and Soil Dynamics, San Diego,
dynamic p-y curves. Can. Geotech. J. 37 (6), 1166–1183.
California, Paper No. 6.27.
Franza, A., Sheil, B., 2021. Pile groups under vertical and inclined eccentric loads:
Mylonakis, G., Gazetas, G., 1999. Lateral Vibration and Internal Forces of Grouped Piles
elastoplastic modelling for performance based design. Comput. Geotech. 135,
in Layered Soil. J. Geotech. Geoenviron. Eng. ASCE 125 (1), 16–25.
104092.
Nova, R., Montrasio, L., 1991. Settlements of shallow foundations on sand. Geotechnique
Gazetas, G., Dobry, R., 1984a. Horizontal response of piles in layered soil. J. Geotech.
41 (2), 243–256.
Eng. Div. 110 (1), 20–40.
Paolucci, R., 1997. Simplified evaluation of earthquake induced permanent
Gazetas, G., Dobry, R., 1984b. Simple radiation damping model for piles and footings.
displacements of shallow foundations. J. Earthq. Eng. 1 (3), 563–579.
J. Geotech. Eng. Div. 110 (6), 937–956.
Rha, C., Taciroglu, E., 2007. Coupled Macroelement Model of Soil-Structure Interaction
Gazetas, G., 1991. Foundation vibrations. Foundation Engineering Handbook, second ed.,
in Deep Foundations. J. Eng. Mech. 133 (12).
Chapter 15, H.-Y. Fang, ed., Chapman and Hall, New York, New York.
Roscoe, K.H., Schofield, A.N., 1956. The stability of short pier foundations on sand. Br.
Gerolymos, N., Papakyriakopoulos, O., Brinkgreve, R.B.J., 2015. Macroelement
Weld. J. 343–354.
modeling of piles in cohesive soil subjected to combined lateral and axial loading.
Proc. 8th Eur. Conf. on Numerical Methods in Geotechnical Engineering, Delft, Balkema,
Rotterdam.

17
D.N. Gorini and L. Callisto Computers and Geotechnics 155 (2023) 105222

Sakellariadis, L., Anastasopoulos, I., 2022. On the Mechanisms Governing the Response Scott, B.D., Park, R., Priestley, M.J.N., 1982. Stress-strain behavior of concrete confined
of Pile Groups Under Combined VHM Loading. Géotechnique. https://doi.org/ by overlapping hoops at low and high strain rates. ACI J. 79, 13–27.
10.1680/jgeot.21.00236. Venanzi, I., Salciarini, D., Tamagnini, C., 2014. The effect of soil-foundation-structure
Salciarini, D., Tamagnini, C., 2009. A hypoplastic macroelement model for shallow interaction on the wind-induced response of tall buildings. Eng. Struct. 79, 117–130.
foundations under monotonic and cyclic loads. Acta Geotech. 4 (3), 163–176. Viggiani, C., Mandolini, A., Russo, G., 2012. Piles and pile groups. Applied Soil
Schanz, T., Vermeer, P.A., Bonnier, P.G., 1999. The hardening-soil model: formulation Mechanics, J. A. Hemsley, Ed., 1st ed. Hoboken, NJ, USA: John Wiley & Sons,
and verification. In: Brinkgreve, R. (ed.), Proceedings of International Symposium: 286–331.
Beyond 2000 in Computational Geotechnics, Amsterdam, the Netherlands. Balkema, Ziegler, H., 1977. An introduction to thermomechanics. North Holland, Amsterdam.
Rotterdam, the Netherlands, pp. 281–290.

18

You might also like